OpenPose(https://github.com/CMU-Perceptual-Computing-Lab/openpose) is one of the most popular pose estimation framework.
There was no problem in installing OpenPose up to the Jetpack 4.4DP version, but when OpenPose was installed in the JetPack 4.4 Production Release, an error occurred. The cause of this problem occurred when some of the CUDA functions used in Caffe were missing. This is explained in detail in Jetson Xavier NX-Human Pose estimation using OpenPose.
However, OpenPose version 1.7.0 released on November 17th, 2020 seems to have solved this problem. If you look at the OpenPose installation documentation, you will find the following contents.
OpenPose uses a custom fork of Caffe (rather than the official Caffe master). Our custom fork is only updated if it works on our machines, but we try to keep it updated with the latest Caffe version. This version works on a newly formatted machine (Ubuntu 16.04 LTS) and in all our machines (CUDA 8 and 10 tested). The default GPU version is the master branch, which it is also compatible with CUDA 10 without changes (official Caffe version might require some changes for it). We also use the OpenCL and CPU tags if their CMake flags are selected. We only modified some Caffe compilation flags and minor details.
Alternatively, you can use your own Caffe distribution on Ubuntu/Mac by 1) disabling
BUILD_CAFFE, 2) settingCaffe_INCLUDE_DIRSto{CAFFE_PATH}/include/caffe, and 3) settingCaffe_LIBSto{CAFFE_PATH}/build/lib/libcaffe.so, as shown in the image below. Note that cuDNN-compatible Caffe version is required in order to get the maximum possible accuracy in OpenPose.
I will install the latest version of OpenPose once again in JetPack 4.5 of the newly installed Jetson Nano.
Prerequisites
For reference, I installed OpenPose while upgrading to OpenCV 4.5. But I think it will be fine with OpenCV 4.1.1 provided by JetPack 4.5.
cmake version check
To build Openpose on the Jetson Nano, you should have cmake version 3.12.2 or higher. First check the cmake version.
cmake --version
If your Jetson Nano's cmake version is lower than 3.12.2, remove the old cmake and rebuild from source codes. Check the latest version at https://github.com/Kitware/CMake/releases . At this time(2021.02.9), Ver 3.19.4 is the latest cmake version. I used the --qt-gui option in bootstrap to build cmake-gui together.
apt-get updateapt-get install -y libssl-dev libcurl4-openssl-dev qt5-default apt-get install -y build-essential libboost-all-dev libboost-dev libhdf5-dev libatlas-base-devapt-get install -y python3-dev python3-pipapt-get remove -y cmake cd /usr/local/src wget https://github.com/Kitware/CMake/releases/download/v3.19.4/cmake-3.19.4.tar.gz tar -xvzf cmake-3.19.4.tar.gz cd cmake-3.19.4 ./bootstrap --qt-gui make -j4 make install
Then you must restart your ssh session. The cmake and cmake-gui are created in the /cmake-3.19.4/bin directory. If the version of cmake or cmake-gui has not been updated in a new ssh or local terminal after make install, copy the cmake and cmake-gui files from the above directory to the /usr/bin directory.
Install OpenPose for JetPack 4.5
Install Protocol Buffer
apt-get install -y libprotobuf-dev protobuf-compiler libgflags-dev libgoogle-glog-dev
Install OpenPose 1.7.0
root@spypiggy-nano:/#cd /usr/local/srcroot@spypiggy-nano:/usr/local/src#wget https://github.com/CMU-Perceptual-Computing-Lab/openpose/archive/v1.7.0.tar.gz root@spypiggy-nano:/usr/local/src#tar -xvzf v1.7.0.tar.gz root@spypiggy-nano:/usr/local/src#cd openpose-1.7.0/ root@spypiggy-nano:/usr/local/src/openpose-1.7.0#cd 3rdparty root@spypiggy-nano:/usr/local/src/openpose-1.7.0/3rdparty#git clone https://github.com/CMU-Perceptual-Computing-Lab/caffe.git root@spypiggy-nano:/usr/local/src/openpose-1.7.0/3rdparty#git clone https://github.com/pybind/pybind11.git
step1. Enable BUILD_PYTHON OPTION
step2. Disable USE_CUDNN option
The use of CUDNN must be excluded in the option. If you build OpenPose 1.7.0 from JetPack 4.5 and you use this option, the build is fine, but phenomena such as system down or segmentation fault(core dump) may occur when running.
step3. Generate Makefile
root@spypiggy-nano:/usr/local/src/openpose-1.7.0/build# make -j4
root@spypiggy-nano:/usr/local/src/openpose-1.7.0/build# make installroot@spypiggy-nano:/usr/local/src/openpose-1.7.0/build# cd pythonroot@spypiggy-nano:/usr/local/src/openpose-1.7.0/build/python# make -j4
Python import Problem
Before you run python samples, first check the openpose python file. Yes it's under the "openpose installation directory/build/python/openpose".
It's time to copy our OpenPose python modules to python3.6 path directory(/usr/lib/python3.6/dist-packages). Then you can import openpose without path problem.
cp -r /usr/local/src/openpose-1.7.0/build/python/openpose/ /usr/lib/python3.6/dist-packages
Let's test your openpose python library.
root@spypiggy-nano:/usr/local/src/openpose-1.7.0/build/python/openpose# python3 Python 3.6.9 (default, Oct 8 2020, 12:12:24) [GCC 8.4.0] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import openpose >>>
Yes, I finally installed the OpenPose 1.7.0 version successfully.
Run a sample program
Let's run a sample program to test whether the OpenPose is properly installed.
root@spypiggy-nano:/usr/local/src/openpose-1.7.0# ./build/examples/openpose/openpose.bin --video ./examples/media/video.avi --net_resolution "256x128" Starting OpenPose demo... Configuring OpenPose... Starting thread(s)... Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
If you see this output, it's a successfull installation.
Be Careful : The Jetson Nano has 4GB of memory. 4GB of memory is not enough to run OpenPose. In the example above, I used the --net_resolution "256x128" option. The reason for using this option is related to memory usage. As the resolution of this option value increases, the processing speed decreases and the memory usage increases. If the resolution is increased above a certain level, the process is forcibly terminated due to insufficient memory.
If you don't use the --net_resolution option, OpenPose will find and use the optimal value. For a 16:9 ratio image, the default value of "656x368" is used. This value must be a multiple of 16.
root@spypiggy-nano:/usr/local/src/openpose-1.7.0# ./build/examples/openpose/openpose.bin --video ./examples/media/video.avi Starting OpenPose demo... Configuring OpenPose... Starting thread(s)... Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0. Killed
<openpose process is killed due to insufficient memory>
JetPack's Ubuntu desktop takes up a lot of memory. Therefore, if you replace the desktop GUI with a lightweight LXDE, you can get an additional 500MB to 1GB of memory. How to acquire additional memory using LXDE is described below:
Wrapping up
Refer to the articles below for how to use OpenPose in the Jetson Nano, Xavier NX, and TX2 series.
- JetsonNano - Human Pose estimation using OpenPose
- Jetson Xavier NX - Human Pose estimation using OpenPose
- JetsonTX2 - Human Pose estimation using OpenPose
If you want to install OpenPose 1.7.0 on the Xavier NX, refer to this article:
In the next article, I'll cover how to use Python with OpenPose 1.7.0.
When I try to run the example I got
답글삭제Error:
Unity plugin only available on Windows.
Coming from:
-/usr/local/src/openpose-1.7.0/src/openpose/utilities/errorAndlog.cpp:DebugInUnity():47
any idea how to solve it?
As you can see from the error message, it seems that unity plugin is only available on Windows 7, 8, 10.
삭제Pls refer the document at https://github.com/CMU-Perceptual-Computing-Lab/openpose_unity_plugin/blob/master/doc/installation.md .